Another look?

From: Harald Nordgren <haraldnordgren@gmail.com>
Date: Jan 2, 2026, 20:27 UTC
Message-ID: <20260102202738.82432-1-haraldnordgren@gmail.com>
In-Reply-To: <6526c419-c60c-49d2-9c1a-52be7aca82f0@gmail.com>

Show 5 quoted lines

> Looking at the tests, it seems that the extra information is shown
> whenever the upstream branch differs from the default push destination
> even if they are on the same remote. I think that is sensible but this
> paragraph should be updated to reflect the fact that one does not need to
> set either of the "pushDefault" config settings to benefit from this.

Good catch, will be included in the next diff!

Show 8 quoted lines

> I'm wondering why we don't reuse the existing messages - I don't see
> how using a different wording for the push destination compared to the
> upstream branch benefits the user. We should adjust the hints that
> are shown so that we only recommend pulling from the upstream branch
> and only recommend pushing to the default push destination but the
> comparisons should be the same. Also I don't find the message "Diverged
> from 'origin/feature' by N commits' very helpful, I'd find it more useful
> if it gave the ahead/behind count like we do for the upstream branch.

Thanks for saying this, I actually did the majority of this work already but was hesitant to include it in the diff because it increases the surface area. But I will include it now!

Show 11 quoted lines

> Shouldn't we be taking account of the push and fetch refspecs here? There
> is no guarantee that $branch maps to refs/remotes/$remote/$branch. To take
> a silly example, if we have
>
>      remote.$remote.fetch =
> refs/heads/$branch:refs/remotes/$remote/abc-$branch remote.$remote.pull =
> refs/heads/$branch:refs/heads/xyz-$branch
>
> Then we should be using "refs/remotes/$remote/abc-xyz-$branch" in the
> message above as "$branch" will be pushed to "xyz-$branch" on the remote
> which is fetched to "$remote/abc-xyz-$branch"

I don't understand this one, can you maybe explain how to code will need to change?

> As we don't use flag we can pass NULL - see the documentation for this
> function in refs.h

Thanks, will be included in the next diff!

> Why are we still suggesting pushing to the upstream branch when we know

the user is pushing to a different remote branch?

Thanks, will be included in the next diff!

Harald